AI guidance reduces exam time by 34% for novice ultrasound operators
Artificial intelligence-enabled guidance during ultrasound procedures could significantly reduce exam times while also improving diagnostic quality.
Ultrasound exams depend highly on operators' skill; the shoulder, in particular, can be difficult for radiologists to accurately assess due to the number or tendons, ligaments and muscles that work together in the area. However, more providers have started utilizing US in recent years to diagnose rotator cuff disorders. As such, it is important for operators to accurately capture images of all the structures critical to the rotator cuff’s function.
A new analysis published in Academic Radiology suggests that an AI tool could help improve the quality and reproducibility of these exams.
“Despite its established diagnostic value, the practical application of shoulder US involves significant complexity. Even for highly experienced MSK radiologists, the process of distinguishing different tendons can be time-consuming, mainly due to the complex anatomy and similar tendons surrounding the humerus within the shoulder,” Hongmei Liu, PhD, with the Affiliated Guangdong Second Provincial General Hospital of Jinan University in China, and colleagues noted. “Thus, it would be exceedingly challenging for inexperienced operators or novices to acquire shoulder US planes proficiently.”
The team developed an AI-enabled tool for the automatic classification and structural recognition of shoulder US planes. Trained on 852 standard plane images and 74,909 frame images from 13,312 shoulder US videos, the tool was designed to accurately guide the acquisition of 15 standard planes and localize 27 key structures.
On external validation, AI achieved an AUC of 0.99 when tasked with guiding operators to the correct structure. It was especially beneficial for junior residents, helping to reduce their examination times by 34% compared to their unassisted passes. The AI's assistance was reportedly comparable to that of expert human radiologists.
Researchers involved in its development suggested it has great potential when used during training.
“Our AI system integrates real-time feedback into the scanning process itself, standardizing plane acquisition and structural positioning of image generation points,” the group noted. “Thus, it serves a dual clinical goal: accelerating training for novices and supporting consistent, high-quality acquisition in daily practice.”
Read more about the tool here.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.